← Back to Agent Etna
Feature
Private
by design.
Your prompts and your agent's data belong to you. Agent Etna was built so privacy is the default — not a tier you upgrade to, not a checkbox you find later.
What "private by design" actually means.
Most platforms treat privacy as something you configure. We treat it as something we never break in the first place. Three concrete commitments, not policies — engineering decisions baked into how the platform works.
01
Your instructions stay with us.
When Agent Etna analyses your agent, it doesn't ship your raw prompt off to an external model. The LLM sees structural information — what kind of agent it is, what the failure looked like, how the architecture is shaped — never the actual instruction text you wrote.
02
Anything Agent Etna remembers is locked.
Per-user memory — preferences, context, notes Agent Etna keeps about how you like to work — is encrypted before it touches disk. Even if someone gets a copy of the storage, they don't get a copy of what's in it. You can read it, edit it, or wipe it any time from Settings.
03
Sharing is off by default.
Some users opt to share their accepted fixes anonymously, so the wider community benefits. The default is off. If you turn it on, you'll see exactly what gets shared, and you can turn it off again at any time. Your data, your call.
What we don't do.
We don't train shared models on your data. We don't sell logs. We don't hold your prompts hostage if you cancel. There's no "premium privacy tier" — every account, free or paid, gets the same defaults.
Auditable, not just claimed.
Privacy is a feature you have to be able to verify. Settings → Privacy shows you what Agent Etna remembers about you, what it shares with whom, every time the audit log records a sensitive action — memory write, consent change, fix applied, change rolled back. You see the same record we see.
Secure by design.
Privacy without security is a press release. Agent Etna's architecture makes the dangerous defaults — autonomous edits, unsigned changes, surprise merges — impossible by construction. Three guarantees baked into how the platform works:
01
Every fix opens a real GitHub PR.
Sandboxed in a labelled branch, run through your existing branch protection, CODEOWNERS, and required reviewers. Nothing reaches your main branch without your team's explicit approval — Agent Etna can't bypass your review process even if we wanted to.
02
Every change carries a tamper-proof signature.
A cryptographic keypair is generated the moment you connect. Every accepted fix, every rolled-back change, every test result is signed and anchored in a public verification log. Anyone can prove a change came from your agent — and detect if it didn't.
03
Human-in-the-loop, by default.
No silent autonomous edits. No black-box merges. Every change to your agent passes through a human checkpoint — your approval, your repository, your audit trail. Even if Agent Etna disappears tomorrow, the proof of every change stays with you.
Privacy you can read,
not just trust.